Psychological Review

Vol. 64, No. 2, 1957

ON PERCEPTUAL READINESS
1
JEROME S. BRUNER
Harvard University
About ten years ago I was party to
the publication of an innocent enough
paper entitled "Value and Need as Or-
ganizing Factors in Perception." It was
concerned with what at that time was
the rather obscure problem of how extra-
stimulus factors influenced perception,
a subject then of interest to only a
small band of us-Gardner Murphy,
Nevitt Sanford, Muzafer Sherif, and a
few others. Obviously, Professor Bor-
ing is quite right about the mischievous-
ness of the Zeitgeist, for the appearance
of this paper seemed to coincide with all
sorts of spirit-like rumblings within the
world of psychology that were soon to
erupt in a most unspirit-like torrent of
research on this very topic-perhaps
three hundred research reports and the-
oretical explications in the ten years
since then. F. H. Allport (1) and M.
D. Vernon (81) have each recently had
a fresh look at the field, sorting out the
findings and evaluating the theoretical
positions, and they have done superb
service. Their labors free me to pursue
a more relaxed course. What I should
like to do in this paper is to set forth
what seem to me to be the outlines of an
approach to perception congruent with
this body of new (and often contradic-
tory) findings and to sketch out what
appear to me to be the persistent prob-
lems still outstanding.
1 The present paper was prepared with the
invaluable assistance of Mr. Michael Wallach.
I also benefitted from the comments of Pro-
fessors W. C. H. Prentice, Karl Pribram, and
M. E. Bitterman, and from various associates
at Princeton University, Kansas University,
and the University of Michigan, where ver-
sions of this paper were presented.
123
ON THE NATURE OF PERCEPTION
Perception involves an act of cate-
gorization. Put in terms of the ante-
cedent and subsequent conditions from
which we make our inferences, we stimu-
late an organism with some appropriate
input and he responds by referring the
input to some class of things or events.
"That is an orange," he states, or he
presses a lever that he has been "tuned"
to press when the object that he "per-
ceives" is an orange. On the basis of
certain defining or criteria} attributes in
the input, what are usually called cues
although they should be called clues
( 35), there is a selective placing of the
input in one category of identity rather
than another. The category need not
be elaborate: "a sound," "a touch," "a
pain," are also examples of categorized
inputs. The use of cues in inferring the
categorial identity of a perceived ob-
ject, most recently treated by Bruner,
Goodnow, and Austin (9) and by Binder
( 4), is as much a feature of perception
as the sensory stuff from which percepts
are made. What is interesting about
the nature of the inference from cue to
identity in perception is that it is in no
sense different from other kinds of cate-
gorial inferences based on defining at-
tributes. "That thing is round and
nubbly in texture and orange in color
and of such-and-such size-therefore an
orange; let me now test its other prop-
erties to be sure." In terms of process,
this course of events is no different from
the more abstract task of looking at a
number, determining that it is divisible
only by itself and unity, and thereupon
categorizing it in the class of prime
numbers. So at the outset, it is evident
124 }EROME S. BRUNER
that one of the principal characteristics
of perceiving is a characteristic of cogni-
tion generally. There is no reason to
assume that the laws governing infer-
ences of this kind are discontinuous as
one moves from perceptual to more con-
ceptual activities. In no sense need the
process be conscious or deliberate. A
theory of perception, we assert, needs a
mechanism capable of infere9ce and
categorizing as much as one is needed in
a theory of cognition.
Let it be plain that no claim is being
made for the utter indistinguishability
of perceptual and more conceptual in-
ferences. In the first place, the former
appear to be notably less docile or re-
versible than the latter. I may know
that the Ames distorted room that looks
so rectangular is indeed distorted, but
unless conflicting cues are put into the
situation, as in experiments to be dis-
cussed later, the room still looks rec-
tangular. So too with such compelling
illusions as the Miller-Lyer: in spite of
knowledge to the contrary, the line with
the extended arrowheads looks longer
than the equal-length one with those in-
clined inward. But these differences,
interesting in themselves, must not lead
us to overlook the common feature of in-
ference underlying so much of cognitive
activity.
Is what we have said a denial of the
classic doctrine of sense-data? Surely,
one may argue (and Hebb [36] has
done so effectively) that there must be
certain forms of primitive organization
within the perceptual field that make
possible the differential use of cues in
identity categorizing. Both logically and
psychologically, the point is evident.
Yet it seems to me foolish and unneces-
sary to assume that the sensory "stuff"
on which higher order categorizations
are based is, if you will, of a different
sensory order than more evolved iden-
tities with which our perceptual world
is normally peopled. To argue other-
wise is to be forced into the contradic-
tions of Locke's distinction between
primary and secondary qualities in per-
ception. The rather bold assumption
that we shall make at the outset is that
all perceptual experience is necessarily
the end product of a categorization
process.
And this for two reasons. The first is
that all perception is generic in the
sense that whatever is perceived is placed
in and achieves its "meaning" from a
class of percepts with which it is
grouped. To be sure, in each thing we
encounter, there is an aspect of unique-
ness, but the uniqueness inheres in devia-
tion from the class to which an object
is "assigned." Analytically, let it be
noted, one may make a distinction, as
Gestalt theorists have, between a pure
stimulus process and the interaction of
that stimulus process with an appropri-
ate memory trace-the latter presumably
resulting in a percept that has an iden-
tity. If indeed there is a "pure stimu-
lus process," it is doubtful indeed that
it is ever represented in perception bereft
of identity characteristics. The phe-
nomenon of a completely unplaceable
object or event or "sensation"-even
unplaceable with respect to modality-
is sufficiently far from experience to be
uncanny. Categorization of an object or
event-placing it or giving it identity-
can be likened to what in set theory
is the placement of an element from a
universe in a subset of that universe of
items on the basis of such ordered di-
mensional pairs, triples, or n-tuples as
man-woman, mesomorph-endomorph-ec-
tomorph, or height to nearest inch. In
short, when one specifies something
more than that an element or object be-
longs to a universe, and that it belongs
in a subset of the universe, one has
categorized the element or object. The
categorization can be as intersecting as
"this is a quartz crystal goblet fashioned
in Denmark," or as simple as "this is a
ON PERCEPTUAL READINESS 125
glassy thing." So long as an operation
assigns an input to a subset, it is an act
of categorization.
More serious, although it is "only a
logical issue," is the question of how one
could communicate or make public the
presence of a nongeneric or completely
unique perceptual experience. Neither
language nor the tuning that one could
give an organism to direct any other
form of overt response could provide
an account, save in generic or categorial
terms. If perceptual experience is ever
had raw, i.e., free of categorial identity,
it is doomed to be a gem serene, locked
in the silence of private experience.
Various writers, among them Gibson
(26), Wallach (83), and Pratt (66),
have proposed that we make a sharp
distinction between the class of per-
ceptual phenomena that have to do with
the identity or object-meaning of things
and the attributive or sensory world
from which we derive our cues for in-
ferring identities. Gibson, like Titch-
ener (78) before him, urges a distinction
between the visual field and the visual
world, the former the world of attribu-
tive sense impressions, the latter of ob-
jects and things and events. Pratt urges
that motivation and set and past ex-
perience may affect the things of the
visual world but not the stuff of the
visual field. And Wallach too reflects
this ancient tradition of his Gestalt fore-
bears by urging the distinction between
a stimulus process pure and the stimulus
process interacting with a memory trace
of past experience with which it has
made a neural contact on the basis of
similarity. The former is the stuff of
perception; the latter the finished per-
cept. From shirtsleeves to shirtsleeves
in three generations: we are back with
the founding and founded content of the
pre-Gestalt Gestalters. If one is to
study the visual field freed of the things
of the visual world, it becomes necessary
-as Wallach implies-to free oneself of
the stimulus error: dealing with a per-
cept not as an object or as a thing with
identity, but as a magnitude or a bright-
ness or a hue or a shape to be matched
against a variable test patch.
If we have implied that categorizing
is often a "silent" or unconscious proc-
ess, that we do not experience a going-
from-no-identity to an arrival-at-iden-
tity, but that the first hallmark of any
perception is some form of identity, this
does not free us of the responsibility of
inquiring into the origin of categories.
Certainly, Hebb (36) is correct in as-
serting like Immanuel Kant, that cer-
tain primitive unities or identities within
perception must be innate or autoch-
thonous and not learned. The primi-
tive capacity to categorize "things"
from "background" is very likely one
such, and so too the capacity to dis-
tinguish events in one modality from
those in others-although the phenom-
ena of synesthesia would suggest that
this is not so complete a juncture as it
might seem; e.g., von Hornbostel (39).
The sound of a buzz saw does rise and
fall phenomenally as one switches room
illumination on and off. The full reper-
tory of innate categories-a favorite
topic for philosophical debate in the
19th century-is a topic on which per-
haps too much ink and too little em-
pirical effort have been spilled. Motion,
causation, intention, identity, equiva-
lence, time, and space, it may be persua-
sively argued, are categories that must
have some primitive counterpart in the
neonate. And it may well be, as Piaget
(65) implies, that certain primitive ca-
pacities to categorize in particular ways
depend upon the existence of still more
primitive ones. To identify something
as having "caused" something else re-
quires, first, the existence of an identity
category such that the two things in-
volved each may conserve identity in the
process of "cause" producing "effect."
Primitive or unlearned categories-a
126 jEROME s. BRUNER
matter of much concern to such students
of instinctive behavior as Lashley (51)
and Tinbergen (77)-remain to be ex-
plicated. In what follows, we shall
rather cavalierly take them for granted.
As to the development of more elaborated
categories in terms of which objects are
placed or identified, it involves the
process of learning how to isolate, weigh,
and use criteria! attribute values, or
cues for grouping objects in equivalence
classes. It is only as mysterious, but
no more so, than the learning of any
differential discrimination, and we shall
have occasion to revisit the problem
later.
A second feature of perception, be-
yond its seemingly categorial and in-
ferential nature, is that it can be de-
scribed as varyingly veridical. This is
what has classically been called the
"representative function" of perception:
what is perceived is somehow a repre-
sentation of the external world-a meta-
physical hodgepodge of a statement but
one which we somehow manage to un-
derstand in spite of its confusion. We
have long since given up simulacra! the-
ories of representation. What we gen-
erally mean when we speak of represen-
tation or veridicality is that perception
is predictive in varying degrees. That
is to say, the object that we see can also
be felt and smelled and there will some-
how be a match or a congruity between
what we see, feel, and smell. Or, to
paraphrase a younger Bertrand Russell,
what we see will turn out to be the same
thing should we take a "closer look" at
it. Or, in still different terms, the cate-
gorial placement of the object leads to
appropriate consequences in terms of
later behavior directed toward the per-
ceived object: it appears as an apple,
and indeed it keeps the doctor away if
consumed once a day.
Let it be said that philosophers, and
notably the pragmatist C. S. Peirce,
have been urging such a view for more
years than psychologists have taken
their urgings seriously. The meaning of
a proposition, as Peirce noted in his
famous essay on the pragmatic theory
of meaning ( 63), is the set of hypo-
thetical statements one can make about
attributes or consequences related to
that proposition. "Let us ask what we
mean by calling a thing hard. Evi-
dently, that it will not be scratched by
many other substances" (White, (84)).
The meaning of a thing, thus, is the
placement of an object in a network of
hypothetical inference concerning its
other observable properties, its effects,
and so on.
All of this suggests, does it not, that
veridicality is not so much a matter of
representation as it is a matter of what
I shall call "model building." In learn-
ing to perceive, we are learning the rela-
tions that exist between the properties
of objects and events that we encounter,
learning appropriate categories and cate-
gory systems, learning to predict and to
check what goes with what. A simple
example illustrates the point. I present
for tachistoscopic recognition two non-
sense words, one a 0-order approxima-
tion to English constructed according to
Shannon's rules, the other a 4-order ap-
proximation: YRULPZOC and VER-
NALIT. At 500 milliseconds of ex-
posure, one perceives correctly and in
their proper place about 48 per cent of
the letters in 0-order words, and about
93 per cent of the letters in 4-order
words. In terms of the amount of in-
formation transmitted by these letter
arrays, i.e., correcting them for redun-
dancy, the subject is actually receiving
the same informational input. The dif-
ference in reportable perception is a
function of the fact that the individual
has learned the transitional probability
model of what goes with what in Eng-
lish writing. We say that perception in
one case is more "veridical" than in the
other-the difference between 93 per
ON PERCEPTUAL READINESS 127
cent correct as contrasted with 48 per
cent. What we mean is that the model
of English with which the individual is
working corresponds to the actual events
that occur in English, and that if the
stimulus input does not conform to the
model, the resulting perception will be
less veridical. Now let us drop the
image of the model and adopt a more
sensible terminology. Perceiving ac-
curately under substandard conditions
consists in being able to refer stimulus
inputs to appropriate coding systems;
where the information is fragmentary,
one reads the missing properties of the
stimulus input from the code to which
part of the input has been referred. If
the coding system applied does not
match the input, what we read off from
the coding system will lead to error and
nonveridical perception. I would pro-
pose that perceptual learning consists
not of making finer and finer discrimi-
nations as the Gibsons (27) would have
us believe, but that it consists rather in
the learning of appropriate modes of
coding the environment in terms of its
object character, connectedness, or re-
dundancy, and then in allocating stimu-
lus inputs to appropriate categorial cod-
ing systems.
The reader will properly ask, as
Prentice (67) has, whether the notion
of perceptual representation set forth
here is appropriate to anything other
than situations where the nature of the
percept is not "clear"-perceptual repre-
sentation under peripheral viewing con-
ditions, in tachistoscopes, under extreme
fatigue. If I am given a very good look
at an object, under full illumination and
with all the viewing time necessary, and
end by calling it an orange, is this a
different process from one in which the
same object is flashed for a millisecond
or two on the periphery of my retina
with poor illumination? In the first
and quite rare case the cues permitting
the identification of the object are super-
abundant and the inferential mechanism
operates with high probability relation-
ships between cues and identities. In
the latter, it is less so. The difference
is of degree. What I am trying to say
is that under any conditions of percep-
tion, what is achieved by the perceiver
is the categorization of an object or
sensory event in terms of more or less
abundant and reliable cues. Repre-
sentation consists of knowing how to
utilize cues with reference to a system
of categories. It also depends upon the
creation of a system of categories-in-
relationship that fit the nature of the
world in which the person must live.
In fine, adequate perceptual representa-
tion involves the learning of appropriate
categories, the learning of cues useful in
placing objects appropriately in such
systems of categories, and the learning
of what objects are likely to occur in
the environment, a matter to which we
will turn later.
We have neglected one important fea-
ture of perceptual representation in our
discussion: representation in perception
of the space-time-intensity conditions of
the external world. Perceptual mag-
nitudes correspond in some degree to
the metrical properties of the physical
world that we infer from the nature of
our perception. That is to say, when
one line looks longer than another, it is
likely to be longer as measured by the
ruler. There are constant errors and
sampling errors in such sensory repre-
sentation, but on the whole there is
enough isomorphism between perceiv-
ing without aids (psychology) and per-
ceiving with aids (physics) to make the
matter perenially interesting.
Is this form of representation subject
to the kinds of considerations we have
been passing in review? Does it depend
upon categorizing activities and upon
the construction of an adequate system
of categories against which stimulus in-
puts can be matched? There is prob-
128 }EROME S. BRUNER
ably one condition where perceptual acts
are relatively free of such influences,
and that is in the task of discriminating
simultaneously presented stimuli as alike
or different-provided we do not count
the "tuning of the organism" that leads
one to base his judgment on one rather
than another feature of the two stimuli.
Ask the person to deal with one stimulus
at a time, to array it in terms of some
magnitude scale, and immediately one
is back in the familiar territory of in-
ferential categorizing. Prentice, in his
able defense of formalism in the study
of perception (67), seems to assume that
there is a special status attached to per-
ceptual research that limits the set of
the observer to simple binary decisions
of "like" and "different" or "present"
and "absent," and to research that also
provides the subject with optimal stimu-
lus conditions, and Graham ( 31) has
recently expressed the credo that no
perceptual laws will be proper or pure
laws unless we reduce perceptual ex-
perimentation to the kinds of operations
used in the method of constant stimuli.
There was at one time a justification
for such a claim on the grounds that
such is the best strategy for getting at
the sensory-physiological processes that
underlie perception. As we shall see in
a later section, current work in neuro-
physiology brings this contention into
serious doubt. In any case, the point
must be made that many of the most
interesting phenomena in sensory per-
ception are precisely those that have
been uncovered by departing from the
rigid purism of the method of constants.
I have in mind such pioneering studies
as those of Stevens on sensory scales,
where the organism is treated as an in-
strument whose sensory categorizations
and scalar orderings are the specific ob-
ject of study (74). Add to this the ad-
vances made by Helson on adaptation
level (37) and by Volkmann on the
anchoring of sensory scales (82)-both
using the "sloppy" method of single
stimuli-and one realizes that the na-
ture of representation in perception of
magnitudes is very much subject to
categorizing processes, and to perceptual
readiness as this is affected by subjective
estimates of the likelihood of occurrence
of sensory events of different magni-
tudes. Indeed, Helson's law of adapta-
tion level states that the subjective mag-
nitude of a singly presented stimulus
depends upon the weighted geometric
mean of the series of stimuli that the
subject has worked with, and the ingeni-
ous experiments of Donald Brown (7)
have indicated that this adaptation level
is influenced only by those stimuli that
the subject considers to be within the
category of objects being considered.
Ask the subject to move a weight from
one side of the table to the other with
the excuse that it is cluttering up the
table, and the weight does not serve as
an anchor to the series, although it will
show a discernible effect if it is directly
included in the series being judged. In
short, the category systems that are
utilized in arraying magnitudes are also
affected by the requirement of match-
ing one's model of the world to the
actual events that are occurring-even
if the categories be no more complicated
than "heavy," "medium," and "light."
The recent work of Stevens (75) on
"the direct estimation of sensory magni-
tudes" highlights the manner in which
veridicality in sensory judgment de-
pends upon the prior learning of an
adequate category set in terms of which
sensory input may be ordered. Sub-
jects are presented a standard tone of
1000 cps at 80 db. sound-pressure-level
and are told that the value of this
loudness is 10. Nine variable loud-
nesses all of the 1000 cps are then pre-
sented, varying 70 db. on either side of
the standard, each one at a time being
paired with the standard. "If the stand-
ard is called 10, what would you call
ON PERCEPTUAL READINESS 129
the variable? Use whatever numbers
seem to you appropriate-fractions,
decimals, or whole numbers." If one
then compares the categorial judgments
made with the sound pressure level of
the various tones presented, using a log-
log plot (log of the magnitude estima-
tion against log of sound-pressure-level),
the resulting function is a straight line,
described by the empirical formula
L = kJO.B,
where Lis loudness and I intensity. In
short, categorial sorting of sensory mag-
nitudes provides one with a mapping or
representation of physical intensity.
There are, to be sure, many problems
connected with such a procedure, but
the point remains: the magnitude cate-
gories in terms of which we scale sensory
events represent a good fit to the physi-
cal characteristics of the world. Call
this "veridicality" if you wish-although
I do not see what is gained thereby;
yet whatever one calls it, one must not
lose sight of the fact that the judgments
made are predictive of other features of
the sensory inputs. Given the empirical
conversion formula, one can predict
from categorial judgment to physical
meter readings.
To summarize, we have proposed that
perception is a process of categorization
in which organisms move inferentially
from cues to categorial identity and that
in many cases, as Helmholtz long ago
suggested, the process is a silent one.
If you will, the inference is often an
"unconscious" one. Moreover, the re-
sults of such categorizations are repre-
sentational in nature: they represent
with varying degrees of predictive ve-
ridicality the nature of the physical
world in which the organism operates.
By predictive veridicality I mean simply
that perceptual categorization of an ob-
ject or event permits one to "go beyond"
the properties of the object or event per-
ceived to a prediction of other properties
of the object not yet tested. The more
adequate the category systems con-
structed for coding environmental events
in this way, the greater the predictive
veridicality that results.
Doubtless, the reader will think of any
number of examples of perceptual phe-
nomena not covered by the simple pic-
ture we have drawn. Yet a great many
of the classic phenomena are covered
-psychophysical judgment, constancy,
perceptual identification, perceptual
learning, and so on. This will become
clearer in the following sections. What
must now be dealt with are the phe-
nomena having to do with selectivity:
attention, set, and the like.
CuE UTILIZATION AND CATEGORY
AccESSIBILITY
A fruitful way of thinking of the na-
ture of perceptual readiness is in terms
of the accessibility of categories for use
in coding or identifying environmental
events. Accessibility is a heuristic con-
cept, and it may be defined in terms of
a set of measures. Conceive of a per-
son who is perceptually ready to en-
counter a certain object, an apple let us
say. How he happens to be in this state
we shall consider later. We measure the
accessibility of the category "apples" by
the amount of stimulus input of a cer-
tain pattern necessary to evoke the per-
ceptual response "there is an apple," or
some other standardized response. We
can state the "minimum" input required
for such categorization by having our
observer operate with two response cate-
gories, "yes" and "no," with the likeli-
hood of occurrence of apples and non-
apples at SO: SO, or by using any other
definition of "maximum readiness" that
one wishes to employ. The greater the
accessibility of a category, (a) the less
the input necessary for categorization
to occur in terms of this category, (b)
the wider the range of input character-
istics that will be "accepted" as fitting
130 jEROME s. BRUNER
the category in question, (c) the more
likely that categories that provide a
better or equally good fit for the input
will be masked. To put it in more
ordinary language; apples will be more
easily and swiftly recognized, a wider
range of things will be identified or mis-
identified as apples, and in consequence
the correct or best fitting identity of
these other inputs will be masked. This
is what is intended by accessibility.
Obviously, categories are not isolated.
One has a category "apples," to be sure,
but it is imbedded by past learning in a
network of categories: "An apple a day
keeps the doctor away" is one such cate-
gory system. So too, are "apples are
fruits" and other placements of an ob-
ject in a general classification scheme.
Predictive systems are of the same order:
e.g., "The apple will rot if not refriger-
ated." We have spoken of these sys-
tems before as the "meaning" of an ob-
ject. We mention them again here to
indicate that though we speak analyti-
cally of separate or isolated categories as
being accessible to inputs, it is quite
obvious that category systems vary in
accessibility as a whole.
It follows from what has just been
said that the most appropriate pattern
of readiness at any given moment would
be that one which would lead on the
average to the most "veridical" guess
about the nature of the world around
one at the moment-best guess here be-
ing construed, of course, as a response
in the absence of the necessary stimulus
input. And it follows from this that the
most ready perceiver would then have
the best chances of estimating situations
most adequately and planning accord-
ingly. It is in this general sense that
the ready perceiver who can proceed
with fairly minimal inputs is also in a
position to use his cognitive readiness
not only for perceiving what is before
him but in foreseeing what is likely to
be before him. We shall return to this
point shortly.
We must turn now to the question of
cue utilization, the "strategies" in terms
of which inferences are made (by the
nervous system, of course) from cue to
category and thence to other cues. I
prefer to use the term strategy for sev-
eral reasons. Perceiving, since it in-
volves inference, rests upon a decision
process, as Brunswik ( 17), Tanner and
Swets (76) and others have pointed
out. Even in the simplest threshold-
measurement test, the subject has the
task of deciding whether what he is
seeing or hearing is noise only or signal-
plus-noise. Given a set of cues, how-
ever presented, my nervous system must
"decide" whether the thing is an air-
plane or a sea gull, a red or a green, or
what not.
There appears, moreover, to be a se-
quence of such decisions involved in
categorizing an object or event. A com-
mon-sense example will make this clear.
I look across to the mantelpiece op-
posite my desk and see a rectangular
object lying on it. If I continue this
pursuit, subsequent decisions are to be
made: is it the block of plastic I pur-
chased for some apparatus or is it a
book? In the dim light it can be either.
I remember that the plastic is down-
stairs in one of the experimental rooms:
the object "is" a book now, and I search
for further cues on its dark red surface.
I see what I think is some gold: it is a
McGraw-Hill book, probably G. A. Mill-
er's Language and Communication that
I had been using late this afternoon. If
you will, the process is a "bracketing"
one, a gradual narrowing of the cate-
gory placement of the object.
Let us attempt to analyze the various
stages in such a decision sequence.
a. Primitive categorization. Before
any more elaborate inferential activity
ca,n QCC\Ir
1
there must be a first, "silent"
ON PERCBPTUAL READINESS 131
process that results in the perceptual
isolation of an object or an event with
certain characteristic qualities. Whether
this is an innate process or one de-
pending upon the prior construction of
a cell-assembly, in the manner of Hebb
( 36), need not concern us. What is re-
quired simply is that an environmental
event has been perceptually isolated and
that the event is marked by certain
spatio-temporal-qualitative characteris-
tics. The event may have no more
"meaning" than that it is an "object," a
"sound," or a "movement."
b. Cue search. In highly practiced
cases or in cases of high cue-category
probability linkage, a second process of
more precise placement based on addi-
tional cues may be equally silent or "un-
conscious." An object is seen with
phenomenal immediacy as a "book" or
an "ash tray." In such instances there
is usually a good fit between the specifi-
cations of a category and the nature of
the cues impinging on the organism-
although "fit" and "probability of link-
age" may stand in a vicarious relation
to each other. Where the fit to accessi-
ble categories is not precise, or when
the linkage between cue and category is
low in probability in the past experience
of the organism, the conscious experi-
ence of cue searching occurs. "What is
that thing?" Here, one is scanning the
environment for data in order to find
cues that permit a more precise place-
ment of the object. Under these cir-
cumstances, the organism is "open" to
maximum stimulation, in a manner de-
scribed below.
c. Confirmation check. When a tenta-
tive categorization has occurred, fol-
lowing cue search, cue search changes.
The "openness" to stimulation decreases
sharply in the sense that now, a tenta-
tive placement of identity having oc-
curred, the search is narrowed for ad-
ditional confirmatory cues to check this
placement. It is this feature of per-
ceptual identification that Woodworth
(85) in his paper on the "Reenforce-
ment of Perception" speaks of as "trial-
and-check." We shall speak of a selec-
tive gating process coming into opera-
tion in this stage, having the effect of
reducing the effective input of stimula-
tion not relevant to the confirmatory
process.
d. Confirmation completion. The last
stage in the process of perceptual identi-
fication is a completion, marked by ter-
mination of cue searching. It is charac-
teristic of this state that openness to
additional cues is drastically reduced,
and incongruent cues are either nor-
malized or "gated out." Experiments
on the perception of incongruity ( 14),
error (69), and the like (15), suggest
that once an object has been categorized
in a high-probability, good-fit category,
the threshold for recognizing cues con-
trary to this categorization increases by
almost an order of magnitude.
The question of fit between cue and
category specification brings us to the
key problem of the nature of categories.
By a category we mean a rule for class-
ing objects as equivalent. The rule
specifies the following about the in-
stances that are to be comprised in the
category.
a. The properties or criterial attribute
values required of an instance to be
coded in a given class.
b. The manner in which such attribute
values are to be combined in making an
inference from properties to category
membership: whether conjunctively (e.g.,
ai and bi), relationally (e.g., ai bears
a certain relation to bt), or disjunctively
(e.g., at or bt).
c. The weight assigned various prop-
erties in making an inference from prop-
erties to category membership.
d. The acceptance limits within which
properties must fall to be criterial. That
132 ]EROME s. BRUNER
is to say, from what range of attribute
values may a,, b, ... kt be drawn.
When we speak of rules, again it should
be made clear that "conscious rules" are
not intended. These are the rules that
govern the operation of a categorizing
mechanism.
The likelihood that a sensory input
will be categorized in terms of a given
category is not only a matter of fit be-
tween sensory input and category speci-
fications. It depends also on the ac-
cessibility of a category. To put the
matter in an oversimplified way, given
a sensory input with equally good fit to
two nonoverlapping categories, the more
accessible of the two categories would
"capture" the input. It is in this sense
that mention was earlier made about
the vicarious relationship between fit
and accessibility.
We have already noted th11t the acces-
sibility of categories reflects the learned
probabilities of occurrence of events in
the person's world. The more frequently
in a given context instances of a given
category occur, the greater the acces-
sibility of the category. Operationally,
this means that less stimulus input will
be required for the instance or event to
be categorized in terms of a frequently
used category. In general, the type of
probability we are referring to is not ab-
solute probability of occurrence, where
each event that occurs is independent
of each other. Such independence is
rare in the environment. Rather, the
principal form of probability learning
affecting category accessibility is the
learning of contingent or transitional
probabilities-the redundant structure
of the environment. That either the
absolute or the contingent probability of
events makes a crucial difference in de-
termining ease of perceptual identifica-
tion is readily supported by research
findings: in the former case by studies
like those of Howes ( 40) and Solomon
and Postman (72), and in the latter by
the work of Miller, Heise, and Lichten
(62) and Miller, Bruner, and Postman
(61).
But the organism to operate ade-
quately must not only be ready for
likely events in the environment, the
better to represent them, and in order
to perceive them quickly and without
undue cognitive strain: it must also be
able to search out unlikely objects and
events essential to its maintenance and
the pursuit of its enterprises. If I am
walking the streets of a strange city and
find myself hungry, I must be able to
look for restaurants regardless of their
likelihood of occurrence in the environ-
ment where I now find myself. In short,
the accessibility of categories I employ
for identifying the objects of the world
around me must not only reflect the en-
vironmental probabilities of objects that
fit these categories, but also reflect the
search requirements imposed by my
needs my ongoing activities, my de-
f e n s e ~ , etc. And for effective search
behavior to occur, the pattern of per-
ceptual readiness during search must be
realistic: tempered by what one is likely
to find in one's perceptual world at that
time and at that place as well as by
what one seeks to find.
Let me summarize our considerations
about the general properties of percep-
tion with a few propositions. The first
is that perception is a decision process.
Whatever the nature of the task set, the
perceiver or his nervous system decides
that a thing perceived is one thing and
not another. A line is longer or shorter
than a standard, a particular object is a
snake and not a fallen branch, the in-
complete word L*VE in the context
MEN L*VE WOMEN is the word
LOVE and not LIVE.
The second proposition is that the
decision process involves the utilization
of discriminatory cues, as do all decision
proc;esses, That is to say, the properties.
ON PERCEPTUAL READINESS 133
of stimulus inputs make it possible to
sort these inputs into categories of best
fit.
Thirdly, the cue utilization process
involves the operation of inference.
Going from cue to an inference of iden-
tity is probably the most ubiquitous and
primitive cognitive activity. The utiliza-
tion of inference presupposes the learn-
ing of environmental probabilities and
invariances relating cues to cues, and
cues to behavioral consequences. Cue
utilization involves various stages: a
primitive step of isolating an object or
event from the flux of environmental
stimulation, stages of cue searching
where the task is to find cues that can
be fitted to available category specifica-
tions, a tentative categorization with
more search for confirming cues, and
final categorization, when cue searching
is severely reduced.
Fourth, a category may be regarded
as a set of specifications regarding what
events will be grouped as equivalent-
rules respecting the nature of criteria}
cues required, the manner of their com-
bining, their inferential weight, and the
acceptance limits of their variability.
Fifth, categories vary in terms of
their accessibility, the readiness with
which a stimulus input with given prop-
erties will be coded or identified in terms
of a category. The relative accessibility
of categories and systems of categories
seems to depend upon two factors: the
expectancies of the person with regard
to the likelihood of events to be en-
countered in the environment; and the
search requirements imposed on the or-
ganism by his needs and his ongoing
enterprises. To use the functionalist's
language, perceptual readiness or acces-
sibility serves two functions: to mini-
mize the surprise value of the environ-
ment by matching category accessibility
to the probabilities of events in the
world about one, and to maximize the
attainment of sought-after objects and
events.
Veridical perception, so our sixth prop-
osition would run, consists of the coding
of stimulus inputs in appropriate cate-
gories such that one may go from cue to
categorial identification, and thence to
the correct inference or prediction of
other properties of the object so cate-
gorized. Thus, veridical perception re-
quires the learning of categories and cate-
gory systems appropriate to the events
and objects with which the person has
commerce in the physical world. When
we speak of the representative function
of perception, we speak of the adequacy
of the categorizing system of the indi-
vidual in permitting him to infer the
nature of events and to go beyond them
to the correct prediction of other events.
Seventh, under less than optimal con-
ditions, perception will be veridical in
the degree to whick the accessibility of
categorizing systems reflects the likeli-
hood of occurrence of the events that
the person will encounter. Where ac-
cessibility of categories reflects environ-
mental probabilities, the organism is in
the position of requiring less stimulus
input, less redundancy of cues for the
appropriate categorization of objects.
In like vein, nonveridical perception
will be systematic rather than random
in its error insofar as it reflects the in-
appropriate readiness of the perceiver.
The more inappropriate the readiness,
the greater the input or redundancy of
cues required for appropriate categori-
zation to occur-where "appropriate"
means that an input is coded in the cate-
gory that yields more adequate subse-
quent predictions.
MECHANISMS MEDIATING PERCEPTUAL
READrNESS
Having considered some of the most
general characteristics of perceiving,
particularly as these relate to the phe-
nomena of perceptual readiness, we must
134 ]EROME s. BRUNER
turn next to a consideration of the kinds
of mechanisms that mediate such phe-
nomena. Four general types of mecha-
nisms will be proposed: grouping and
integration, access ordering, match-mis-
match signaling, and gating. They will
be described in such a form that they
may be considered as prototypes of
neural mechanisms and, where possible,
neurophysiological counterparts will be
described briefly. Six years ago, Ed-
ward Tolman (79) proposed that the
time was perhaps ripe for reconsidering
the neural substrate of perception. Per-
haps he was right, or perhaps even now
the enterprise is somewhat premature.
Yet, the body of perceptual data avail-
able makes it worth while to consider
the kinds of mechanisms that will be
required to deal with them. To use
Hebb's engaging metaphor, it is worth
while to build a bridge between neuro-
physiology and psychology provided we
are anchored at both ends, even if the
middle of the bridge is very shaky.
Grouping and Integration
It is with the neural basis of the cate-
gorizing process that Hebb's Organiza-
tion o j Behavior ( 36) is principally con-
cerned. Little is served by recapitulat-
ing his proposals here, for the reader
will be familiar with the concise ac-
count in Chapters 4 and 5 of that book,
where the concepts of cell assembly and
phase sequence are set forth with a
clarity that permits one to distinguish
what is neurophysiological fact and
what speculation. In essence, Hebb's
account attempts to provide an anatomi-
cal-physiological theory of how it is that
we distinguish classes of events in the
environment, and how we come to rec-
ognize new events as exemplars of the
once established classes. The theory
seeks also to provide a mechanism for
integration of sorting activity over time:
the formation of phase sequences for
the conservation of superordinate classes
of events and superordinate sequences.
Basically, it is an associational or an
"enrichment" theory of perception at
the neural level, requiring that estab-
lished neural associations per-
ception of events that have gone to-
gether before. The expectancies, the
centrally induced facilitations that oc-
cur prior to the sensory process for
which they are appropriate, are learned
expectancies based on the existence of
frequency integrators. These frequency
integrators may be neuroanatomical in
the form of synaptic knobs, or they may
be any process that has the effect of
making activity in one locus of the
brain increase or decrease the likelihood
of activity in another. To be sure,
Hebb's theory depends upon some broad
assumptions about convergence of firing
from area 17 outward, about synchro-
nization of impulses, and about the
manner in which reverberatory circuits
can carry organization until the much
slower process of anatomical change can
take place. But this is minor in com-
parison with the stimulation provided
by facing squarely the question of how
the known facts of categorization and
superordination in perception could be
represented in the light of present knowl-
edge.
While it is difficult indeed to propose
a plausible neural mediator to account
for category formation and the develop-
ment of elaborated categorial systems
(e.g., our knowledge of the relations be-
tween classes of events in the physical
world which we manipulate in everyday
life), it is less difficult to specify what
such mechanisms must account for in
perceptual behavior.
At the level of the individual category
or cell assembly, the phenomena of ob-
ject identity must be accounted for.
Moreover, identity conservation or ob-
ject constancy requires explanation in
terms common with the explanation of
identity. Experiments by Piaget (65)
ON PERCEPTUAL READINESS 135
suggest that the capacity to maintain
the phenomenal identity of an object
undergoing change is the hard-won re-
sult of maturation-and-learning. In
connection with the later discussion of
gating processes, we shall have occa-
sion to consider the manner in which,
at different stages in cue utilization, the
required fit between an input and a cell
assembly changes.
Where integration is concerned, there
must be a process capable of conserving
a record of the likely transitions and
contingencies of the environment. The
moment-to-moment programming of per-
ceptual readiness depends upon such
integrations. In short, the relation be-
tween classes of events is conserved in
such a way as to be subject to change
by learning. Several things can be
guessed about integration processes. It
is unlikely that it is a simple autocor-
relation device. Clearly, the concep-
tions of transitional probabilities that
are established in dealing with sequences
of events show biases that no self-re-
specting autocorrelation computer would
be likely to operate with. One of these
is a strong and early tendency to treat
events as nonindependent of each other
over time. In the absence of evidence,
or even in the presence of contrary evi-
dence, humans-as their behavior has
been observed in choice tasks, e.g., Estes
( 23), Goodnow ( 29) -treat random
sequences of events as though they were
governed by dependent probabilities.
The spate of research on two-choice de-
cision behavior has made us quite sharply
aware of this characteristic of cognitive
functioning. The typical pattern is the
gambler's fallacy or, more properly, the
negative recency effect. Given two equi-
probable events whose occurrences are
random, the repetition of one event pro-
gressively leads to the expectancy of
the other. As in the elegant experi-
ments of Jarvik (44) and Goodnow
(29), the probability that a person will
predict one of two events increases di-
rectly as a function of the number of
repetitions of the other event. Such be-
havior persists over thousands of op-
portunities for teiting, and it appears
under a variety of testing conditions
(9).
The second feature of sequential prob-
ability integration mechanisms is that,
in establishing a conception of the prob-
ability with which events will occur, the
typical human subject will bias his esti-
mate in terms of desired or feared out-
comes. As in the experiments of Marks
(60) on children and of Irwin (41) on
adults, the subjectively estimated prob-
ability of strongly desired events will be
higher per previous encountered occur-
rence than the estimated likelihood of
less desired events. Quite clearly, then,
the establishment of estimates depends
upon more than frequency integrations
biased by assumptions of nonindepend-
ence. The "something more" is a moti-
vational or personality process, and we
shall have more to say about it in con-
sidering phenomena of so-called "per-
ceptual sensitization" and "perceptual
defense."
Access Ordering
The term "accessibility" has been
used in preceding pages to denote the
ease or speed with which a given stimu-
lus input is coded in terms of a given
category under varying conditions of in-
struction, past learning, motivation, etc.
It has been suggested, moreover, that
two general sets of conditions affect ac-
cessibility: subjective probability esti-
mates of the likelihood of a given event,
and certain kinds of search sets induced
by needs and by a variety of other
factors.
Let us consider a few relevant facts
about perception. The first of these is
that the threshold of recognition for
stimuli presented by visual, auditory, or
other means is not only a function of
136 }EROME s. BRUNER
the time, intensity, or of
the stimulus input, but also varies mas-
sively as a function of the number of
alternatives for which the perceiver is
set. The size of the expected array, to
say it another way, increases the iden-
tification threshold for any item in the
array. Typical examples of this gen-
eral finding are contained in papers by
Miller, Heise, and Lichten (62) and by
Bruner, Miller, and Zimmerman (10).
The actual shape of the function need
not concern us, save that it is quite clear
that it is not what one would expect
from a simple binary system with a fixed
channel capacity. What we are saying
holds, of course, only for the case where
the perceiver has learned that all the
items in the expected array are (a)
equiprobable and (b) independent, one
of the other, in order of appearance.
The first hunch we may propose,
then, about access-ordering mechanisms
is that degree of accessibility of coding
categories to stimulus inputs is related
to regulation of the number of preac-
tivated cell assemblies that are operative
at the time of input. In an earlier
paper (8), discussing factors that
strengthen an hypothesis in the sense of
making it more easily confirmable, I
proposed that one of the major deter-
minants of such strength was monopoly:
where one and only one hypothesis is
operative with no competing alterna-
tives, it tends to be more readily con-
firmable. It is the same general point
that is being made here. Accessibility,
then, must have something to do with
the resolution of competing alternatives.
As between two arrays of expected al-
ternatives, each of the same size, we may
distinguish between them in terms of the
bias that exists in terms of expected like-
lihood of occurrence of each alternative.
If one could characterize the expected al-
ternatives in terms of probability values,
one could conceive of the array ranging
in values from a figure approaching 1.0 at
one extreme, to another approaching 0.0
at the other. The findings with respect
to perceptual readiness for the alterna-
tives represented in such an array are
well known. For a constant-sized ar-
ray, the greater the estimated likelihood
of occurrence of an alternative, the more
readily will the alternative be perceived
or identified. This is known to be true
for large arrays, such as the ensemble of
known words in the English language,
whose likelihood may be roughly judged
by their frequency of occurrence in
printed English (e.g., 40). It is not
altogether clear that it is the case for
arrays of expected alternatives that are
within the so-called span of attention-
i.e., less than seven or eight alternatives.
That the principle holds for middling
arrays of about 20 items has been shown
by Solomon and Postman (72).
What is particularly interesting about
change of accessibility, under conditions
where estimates of the likelihood of oc-
currence of alternatives become biased,
is that the biasing can be produced
either by a gradual learning process akin
to probability learning or by instruction.
Thus, Bitterman and Kniffin ( 5), in-
vestigating recognition thresholds for
taboo and neutral words, show that as
the experiment progresses, there is a
gradual lowering of threshold for the
taboo words as the subject comes to
expect their occurrence. Bruner and
Postman ( 14) have similarly shown
that repeated presentation of stimulus
materials containing very low-probability
incongruities leads to a marked decrease
in threshold time required for recog-
nizing the incongruous features. At the
same time, both Cowen and Beier (20)
and Postman and Crutchfield (70) have
shown that if a subject is forewarned
that taboo words are going to be pre-
sented, his threshold for them will tend
to be lower than for neutral words,
whereas it will be higher if no instruc-
tion is given. In short, preactivation of
ON PERCEPTUAL READINESS 137
cell assemblies-assuming for a moment
that degree of preactivation is the mech-
anism that represents subjective esti-
mates of likelihood of occurence of an
event-such preactivation can be pro-
duced by gradual learning or quantally
by instruction. Moreover, biasing may
be produced by the nature of the situa-
tion in which the perceiver is operating.
A recent study by Bruner and Minturn
( 11) illustrates the point. Subjects are
presented at brief exposure a broken
capital B with a small separation be-
tween the vertical and the curved com-
ponent of the letter so that it may be
perceived as a B or as a 13. The man-
ner in which it is reported is determined
by whether the subject has previously
been presented with letters or with num-
bers to recognize. In short, expectancy
of one or the other context preactivates
a related array of categories or cell-as-
semblies, not just a single, isolated one.
What the neural correlates of access
ordering will look like is anybody's
guess. Lashley (52) has remarked that,
for all our searching, we have not lo-
cated a specific memory trace-either
in the form of a reverberatory circuit,
a definite change in fiber size as pro-
posed by J. Z. Young (88) and Eccles
(21), a synaptic knob-in the manner
of Lorente de No (57) or in any known
form. To be sure, Penfield ( 64) has
activated memories by punctate elec-
trical stimulation of the cortex, but this
is a long remove from a definition of the
neural properties of the trace. For the
time being, one does better to deal in
terms of the formal properties that a
trace system must exhibit than to rest
one's psychological model on any neuro-
physiological or anatomical conception
of the memory trace.
And, quite clearly, one of the formal
properties of a trace system is that its
elements vary in accessibility to stimu-
lus input with the kinds of conditions we
have considered. It is instructive to
note that when a theory of traces lacks
this feature, it ceases to be useful in
dealing with the wide range of per-
ceptual categorizing phenomena of which
we now have knowledge. Gestalt theory
is a case in point. According to Kohler's
view ( 48), a stimulus process "finds"
its appropriate memory trace, resulting
in identification of the stimulus process,
on the basis of distinctive similarity be-
tween stimulus process and memory
trace. The theory has been criticized,
justly I think, for failing to specify the
nature of this similarity save by saying
that it is a neural isomorph of phe-
nomenal similarity. But since similarity
may be highly selective-two objects
may be alike in color but differ in
dozens of other respects-there is ob-
viously some tertium quid that deter-
mines the basis of similarity. More
serious still is the inability of such a
theory to deal with the increased likeli-
hood of categorization in terms of par-
ticular traces as a function of changes
in search set or subjective likelihood
estimates. The Bruner-Minturn results
would require that, as between two
traces with which a stimulus process
may make contact, each equally "simi-
lar" to the stimulus, the stimulus process
will make contact with the one having
a higher probability of being matched by
environmental events. This is interest-
ing, but it is far from the spirit of
Gestalt theory.
Match-Mismatch Processes
One may readily conceive of and, in-
deed, build an apparatus that will ac-
cept or reject inputs on the basis of
whether or not they fulfill certain speci-
fications. Selfridge ( 71) has constructed
a machine to read letters, Fry (24) has
one that will discriminate various pho-
nemes, and Uttley (80) has constructed
one that, like Tinbergen's graylay geese,
will recognize the flying silhouette of a
predator hawk. All such machines have
138 }EROME s. BRUNER
in common that they require a match
between a stimulus input and various
specifications required by the sorting
mechanism of the machine.
In the examples just given, there is no
consequence generated by whether a
given input fulfills the specifications re-
quired by the identifying machine. It
fits or it doesn't fit. But now let us
build in two other features. The first
is that the machine emit a signal to in-
dicate how closely any given input comes
to fulfilling the specifications required:
either by indicating how many attributes
the object has in common with the
specifications, or by indicating how far
off the mark on any given attribute di-
mension a given input is. The second
is that the machine do something on
the basis of these signals: to increase
sensitivity if an object is within a given
distance of specifications for a closer
look, or to decrease it if the object is
further than a certain amount from
specifications, or to stop registering fur-
ther if the input fits.
In short, one can imagine a nervous
system that emits aU-or-none match-
mismatch signals or graded match-mis-
match signals, and one can also imagine
that these signals could then feed into
an effector system to regulate activity
relevant to continuing search behavior
for a fitting object, or to regulate other
forms of activity. MacKay (59) has
recently proposed such a model.
We must return for a moment to an
earlier discussion. In the discussion of
cue utilization, a distinction was made
between three phases of "openness" in
cue search. The first was one in which
a given input was being scanned for its
properties so as to place it in one of a
relatively large set of possible alterna-
tive categories. Here one would register
on as many features of an object as
possible. In a second stage, the input
has been tentatively placed, and the
search is limited to confirming or in-
firming criteria! cues. Finally, with
more definite placement, cue search is
suspended and deviations from specifica-
tion may even be "normalized." It is
for the regulation of such patterns of
search or cue utilization that some mech-
anism such as match-mismatch signal-
ing is postulated.
Let it be said that while match-
mismatch signaling-effector systems are
readily conceivable and readily con-
structed, there is no knowledge available
as to how a system like the nervous sys-
tem might effect such a process. That
there is feedback all over the system is
quite apparent from its detailed anat-
omy, and this is the process out of
which a larger-scale system such as we
have described would be constructed.
Gating Processes
The picture thus far presented is of
a conceptual nervous system with a mas-
sive afferent intake that manages some-
how to sort inputs into appropriate as-
semblies of varying accessibility. It
seems unlikely that this is the nature of
the nervous system, that there should
be no gating or monitoring of stimulus
input short of what occurs at higher
centers. It is with this more peripheral
form of screening of inputs that we shall
now be concerned.
It has long been known that the con-
cept of the "adequate stimulus" could
not simply be defined as a change in en-
vironmental energy sufficient to stimu-
late a receptor. For quite evidently, a
stimulus could be peripherally adequate
in this sense and not be "centrally" ade-
quate at all, either in eliciting electrical
activity in the cortex or in producing a
verbal report of a change in experience
by the subject. Indeed, the very nature
of such complex receptor surfaces as the
retina argues against such a simple no-
tion of "adequate stimulus." For the
reactivity of even a retinal cell at the
fovea seems to be "gated" by the state
ON PERCEPTUAL READINESS 139
of stimulation of neighboring cells.
Thus, if cells A, B, and C lie next each
other in that order in a row, stimulation
of B suppresses the sensitivity of C. If
A now be stimulated, B is suppressed
and C is released or heightened in sensi-
tivity. So even at the level of the first
synapse of a sensory system, there is
mediation outward or gating from in-
ternuncial to receptor cells that pro-
grams the nature of the input that can
come into the sensory system. And to
be sure, there are many phenomena in
perception itself that speak for this same
kind of gating. When we are fixated
upon the vase in the Rubin reversible
figure, the background recedes, is less
surfacy, and in general seems to provide
a generally less centrally adequate form
of sensory input. So too with the studies
of Yokoyama (87) and Chapman (19)
where subjects, set to report on one of
several attributes of briefly presented
stimuli, accomplished their selective task
with a loss of ability to discriminate on
the attributes for which they had not
been set. We shall propose that such
phenomena are very likely mediated by
a gating process which "filters" input
before ever it reaches the cortex.
There is now a growing body of neuro-
physiological evidence that part of this
screening process is relegated to periph-
eral levels of the nervous system-even
as far out as the second synapse of spe-
cialized sensory systems. In an earlier
paper I used the rather fanciful phrase
that "perception acts sometimes as a
welcoming committee and sometimes as
a screening committee." It now appears
that both these committees are closer to
the entrance port than previously con-
ceived.
Consider first the evidence of Kuffler
and Hunt (50) on so simple a "reflex"
as the stretch reflex of the biceps femoris
muscle of the cat in an isolated spinal
nerve-muscle preparation. Recall a lit-
tle anatomy first. Muscle tissue con-
tains special cells called spindles that
are receptors in function, discharging
with contraction or stretch of the muscle
in which they are imbedded. The mus-
cle itself is innervated by an efferent
nerve trunk emerging from the ventral
horn of the spinal cord and, in turn,
an afferent nerve travels to the dorsal
root of the spinal cord. According to
the classical law of Bell and Magendie,
the ventral root of the spinal cord car-
ries efferent-motor impulses down to the
muscles, while the dorsal root carries
sensory impulses up to the cord. Now,
it has been known for a long time that
the presumed efferent nerve going to
muscles carries fibers of large and of
small diameter. A quarter-century ago
Eccles and Sherrington showed that the
ventral nerve branch supplying the bi-
ceps femoris of the cat shows a "strik-
ing division of the fibers into two diam-
eter groups" ( 49), one group centering
around Sp.. in diameter, the other around
15 or 16p... The large fibers are, of
course, fast conductors, the small ones
slow. Leksell (55) has shown that stim-
ulation of the slow-conducting smaller
fibers did not cause detectable contrac-
tions or propagated muscle impulses.
Whtn the larger and fast-conducting
fibers are stimulated, the usual motor-
unit twitch occurred. Kuffler and Hunt
(50) state that, in the lumbosacral out-
flow, about % of the fibers are of the
large-diameter, fast-conduction type; the
other third are of the small type that in
mammalia are "ineffective in directly
setting up significant muscular contrac-
tion." There has been much speculation
about what these fibers are there for,
and the answer is now fairly clear. It
is revolutionary in its implications and
brings deeply into question both the
classical Bell-Magendie law and the
simplistic notion of the reflex arc on
which so much of American learning
theory is based.
It is this. The small fibers of the
140 jEROME s. BRUNER
presumably motor trunk go to the spin-
dle cells and the activity in these fibers
serve to modulate or gate the receptivity
of these specialized sensory endings.
For example, if the small-diameter fibers
are firing into the muscle spindle it may
speed up the amount of firing from this
cell into the afferent nerve that is pro-
duced by a given amount of stretch ten-
sion on the muscle. We need not go
into detail here. It suffices to note that
the state of presumed motor discharge
does not simply innervate the muscle; it
also regulates the amount and kind of
kinesthetic sensory discharge that the
sensory cells in the muscle will send
back to the central nervous system. In-
stead of thinking of a stimulus-response
reflex arc, it becomes necessary even at
this peripheral level to think of the ef-
ferent portion of the arc acting back on
sensory receptors to change the nature
of the stimulus that can get through.
Two additional pieces of evidence on
gating mechanisms at higher levels of
integration may be cited. Where vision
is concerned, Granit (32) has recently
shown that pupillary changes produced
by the ciliary muscle of the eye create
changes in the pattern of firing of the
retina: changes in muscular state work-
ing its way back through the nervous
system into the visual system and back
outward to the retina. There is also
evidence of gating working from the
visual system backward in the opposite
direction: during binocular rivalry, the
nondominant eye shows a less sensitive
pupillary reflex than the dominant eye.
Finally, we may cite the recent evi-
dence of Hernandez-Peon, Scherrer, and
Jouvet (38) working in Magoun's lab-
oratory, work confirmed by analogous
findings of Golambos, Sheatz, and Ver-
nier (28) at the Walter Reed Hospital.
If one stimulates the cat with auditory
clicks, it is possible to record an evoked
spike potential from the cochlear nu-
cleus. Repetition of the clicks leads
to a gradual diminution of the evoked
potential, as if the organism were adapt-
ing. It is quite extraordinary that such
adaptation should be registered as far
out peripherally as the cochlear nucleus,
which is, after all, only the second
synapse of the VIIIth nerve. Now, if
the clicks are previously used as condi-
tioned stimuli signaling shock, the dimi-
nution of the evoked potential no longer
occurs upon repetition of the clicks.
Evidence that the response from the
brain is not being produced by the mus-
cular activity produced by the click as
a conditioned stimulus is provided by
the fact that the same kind of effects are
obtained from cats with temporarily in-
duced muscular paralysis. Further, if
one take a cat whose cochlear nucleus is
still firing upon click stimulation and
introduce a mouse into its visual field,
the clicks no longer evoke a spike po-
tential. A fish odor or a shock to the
paw has the same effect of inhibiting
spike potentials at the cochlear nucleus,
if these distracting stimuli occur con-
currently with the click. "Distraction"
or "shifting of attention" appears to
work its way outward to the cochlear
nucleus.
2
Perhaps the foregoing account has
been needlessly detailed on the side of
neurophysiology. Yet, the interesting
implications of the findings for per-
ceptual theory make such an excursion
worth while. That the nervous system
accomplishes something like gating is
quite clear, even without the neurophysi-
ological evidence. The data of behavior
are full of examples, and the phenomena
of attention require some such mecha-
nism to be explained. Indeed, it is quite
2
Since the above was written, evidence has
been presented by Golambos indicating that
efferently controlled inhibition operates as far
out to the periphery as the hair cells of the
organ of Corti and fibers carrying such in-
hibitory impulses have been traced as far cen-
trally as the superior olivary nucleus-not very
far, but a start.
ON PERCEPTUAL READINESS 141
clear that the nervous system must be
capable of more selective gating than
physiology has yet been able to discover.
That is to say, there must be a filter
somewhere in the eat's nervous system
that will "pass" the squeak of the mouse
in the Hernandez-Peon experiment but
not the cough of the experimenter. And
it is to this problem that we turn now.
I would propose that one of the mech-
anisms operative in regulating search
behavior is some sort of gating or filter-
ing system. In the preceding section, it
was proposed that the "openness" of the
first stage of cue utilization, the "selec-
tivity" of the second stage, and the
"closedness" of the third stage were
probably regulated by a match-mis-
match mechanism. What may be pro-
posed here is that the degree of "open-
ness" or "closedness" to sensory input
during different phases of cue utilization
is likely effected by the kind of gating
processes we have been considering.
How these work in intimate detail is far
from known, yet the work of the last
years in neurophysiology suggests that
we are drawing closer to an answer.
Having considered some general prop-
erties of perception and some possible
mechanisms underlying these, we turn
now to some selected problems in per-
ception better to explore the implica-
tions of what has thus far been pro-
posed.
ON FAILURE OF READINESS
From the foregoing discussion, it is
clear that veridical perception under
viewing or listening conditions that are
less than ideal depends upon a state of
perceptual readiness that matches the
probability of occurrence of events in
the world of the perceiver. This is true,
of course, only in a statistical sense.
What is most likely to occur is not
necessarily what will occur, and the per-
ceiver whose readiness is well matched
to the likelihoods of his environment
may be duped. In Farquhar's hand-
some seventeenth-century phrase: "I
cou'd be mighty foolish, and fancy my-
self mighty witty; reason still keeps its
Throne--but it nods a little, that's all."
The only assurance against the nodding
of reason or probability, under the cir-
cumstances, is the maintenance of a
flexibility of readiness: an ability to per-
mit one's hypotheses about what it is
that is to be perceptually encountered
to be easily infirmed by sensory input.
But this is a topic for later.
There appear to be two antidotes to
nonveridical perception, two ways of
overcoming inappropriate perceptual
readinesses. The one is a re-education
of the misperceiver's expectancies con-
cerning the events he is to encounter.
The other is the "constant close look."
If the re-education succeeds in produc-
ing a better match between internal ex-
pectancies and external event-proba-
bilities, the danger of misperception
under hurried or substandard conditions
of perceiving is lessened. But the mat-
ter of re-educating perceptual expect-
ancies is complex. For where conse-
quences are grave, expectancy concern-
ing what may be encountered does not
change easily, even with continued op-
portunity to test the environment. In
this concluding section we shall consider
some of the factors that contribute to
states of perceptual "unreadiness" that
either fail to match the likelihood of en-
vironmental events or fail to reflect the
requirements of adjustment or both.
Before turning to this task, a word is
in order about the "constant close look"
as an antidote to inappropriate percep-
tual readiness. There is for every cate-
gory of objects that has been estab-
lished in the organism a stimulus input
of sufficient duration and cue redundancy
such that, if the stimulus input fits the
specifications of the category, it will
eventually be correctly perceived as an
exemplar of that category. With enough
142 jEROME s. BRUNER
time and enough testing of defining cues,
such "best fit" perceiving can be ac-
complished for most but not all classes
of environmental events with which the
person has contact. There are some
objects whose cues to identity are suf-
ficiently equivocal so that no such reso-
lution can be achieved, and these are
mostly in the sphere of so-called in-
terpersonal perception: perceiving the
states of other people, their character-
istics, intentions, etc., on the basis of
external signs. And since this is the
domain where misperception can have
the most chronic if not the most acute
consequences, it is doubtful whether a
therapeutic regimen of "close looking"
will aid the misperceiver much in deal-
ing with more complex cue patterns.
But the greatest difficulty rests in the
fact that the cost of close looks is gen-
erally too high under the conditions of
speed, risk, and limited capacity im-
posed upon organisms by their environ-
ment or their constitutions. The ability
to use minimal cues quickly in cate-
gorizing the events of the environment
is what gives the organism its lead time
in adjusting to events. Pause and close
inspection inevitably cut down on this
precious interval for adjustment.
Inappropriate Categories
Perhaps the most primitive form of
perceptual unreadiness for dealing with
a particular environment is the case in
which the perceiver has a set of cate-
gories that are inappropriate for ade-
quate prediction of his environment. A
frequently cited example of such a case
is Bartlett's account ( 3) of the African
visitors in London who perceived the
London bobbies as especially friendly
because they frequently raised their right
hand, palm forward, to the approach-
ing traffic. The cue-category inference
was, of course, incorrect, and they should
have identified the cue as a signal for
stopping traffic. The example, however,
is not particularly interesting because
it is a transient phenomenon, soon cor-
rected by instruction.
A more interesting example, because
it is far less tractable, is provided by
second-language learning and the learn-
ing of a new phonemic system. Why
is it, we may ask, that a person can
learn the structure of a new language,
its form classes, morphemes, lexemes,
and so on, but still retain a "foreign
accent" which he cannot, after a while,
distinguish from the speech flow of na-
tive speakers around him? And why is
it that a person learning a new language
can follow the speech of a person with
his own kind of foreign accent more
readily than he can follow a native
speaker? The answer lies, I think, in
the phenomenon of postcategorization
sensory gating: once an utterance has
been "understood" or decoded in ap-
propriate categories, on the basis of some
of the diacritica of the speech flow, the
remaining features are assimilated or
normalized or screened out. The pho-
nemic categories that are used, more-
over, are modifications of those in the
first language of the speaker. Normal-
ization is in the direction of these first-
language phonemic categories. It is
only by a special effort that, after hav-
ing achieved adequate comprehension of
the second language, one can remain
sensorially "open" enough to register
on the deviation between his own pho-
nemic pattern and that of native speak-
ers. And since there is common cate-
gorization of the "meaning" of utter-
ances by the native speaker and the
fluent foreigner, there is no built-in in-
centive for the foreigner to maintain a
cognitively strainful regimen of attend-
ing further to speech sounds.
Lenneberg (56) has recently shown
the difficulties involved in learning new
modes of categorizing such continua as
chromatic colors. He taught subjects
various nonsense languages, explaining
ON PERCEPTUAL READINESS 143
to them that the words were Hopi names
for colors and that their task was to
learn what colors they stood for. His
stimulus materials were graded Munsell
colors going in a circle from brown,
through green, through blue, through
pink, and then back to brown. A stand-
ardizing group was used to find the
frequency distribution of color naming
over the circle when the English color
names mentioned above were used. Ex-
perimental groups, six in number, were
then run, each being exposed to the use
of the nonsense color names "as these
are used by the Hopi." Then they were
tested on their usage of the names. A
first group was taught the nonsense words
with exact correspondence to the usage
found for the standardizing group on
brown, blue, green, and pink. The other
groups were given distorted usage train-
ing-distorted from English usage. The
distortions were both in the slopes of the
frequency of usage and in the points on
the color continua where the highest
usage frequencies fell. That is to say,
the mode of a distribution in some
cases would fall at a color which in Eng-
lish had no specific name, or fall be-
tween two English categories.
The principal results of the experi-
ment are these. If the reference and
probability relationship is the same for
a nonsense language as it is for English,
relearning is very rapid. The slightest
deviation from this correspondence in-
creases difficulty of learning quite mark-
edly. It is disturbing either to shift
the center of the categories on the color
continuum or to change the shape of
the frequency-of-calling functions, even
when these are made more determina-
tive (i.e., rectilinear) than they nor-
mally are. A shift in the shape of the
frequency-of-calling functions is more
disruptive than a shift in placement on
the color continuum. What is quite
striking is that a highly determinative
frequency-of-calling function can be
learned much more rapidly than one in
which there is a gradual transition in
color naming from one color to another
on the color continuum.
Now, I suspect that the difficulty in
learning a set of neighboring categories
with a state of equivocality prevailing in
the area between the "typical instances"
of each category comes precisely from
the tendency to normalize in the direc-
tion of the center of one category or the
other. If there is a sharp transition be-
tween one color category and another,
this tendency aids learning. If the
transition is gradual, it hinders it. For
it is noteworthy, as in the experiment
of Bruner, Postman, and Rodrigues (16)
that equivocal colors are readily sub-
ject to assimilation in the direction of
expected value.
It is perhaps in the realm of social
perception, where the problem of vali-
dating one's categorizations is severe,
that one finds the most striking effects
of inappropriate category systems. What
is meant here by validation is the test-
ing of the predictions inherent in a
categorization. If, on the basis of a few
cues of personal appearance, for ex-
ample, one categorizes another person
as dishonest, it is extremely difficult in
most cases to check for the other cues
that one would predict might be asso-
ciated with instances of this category.
There is either a delay or an absence of
opportunity for additional cue checking.
Moreover, there is also the likelihood,
since cues themselves are so equivocal in
such a case, that available equivocal
signs will be distorted in such a manner
as to confirm the first impression. It is
much as in the experiments of Asch (2)
and of Haire and Grunes ( 33) on the
formation of first impressions, where
later cues encountered are cognitively
transformed so as to support the first
impression. The reticence of the man
we categorize as dishonest is seen as
"caginess; " the "honest" man's reticence
144 }EROME s. BRUNER
is seen as "integrity" and "good judg-
ment."
It is perhaps because of this difficulty
of infirming such categorial judgments
that an inappropriate category system
can be so hard to change. The slum boy
who rises to the top in science can
change his categories for coding the
events of the physical world quite read-
ily. He has much more difficulty in
altering the socially related category
system with which he codes the phe-
nomena of the social world around him.
Inappropriate Accessibility Ordering
Perhaps the most noticeable "percep-
tual unreadiness" comes from interfer-
ence with good probability learning by
wishes and fears. I have in mind the
kind of distorted expectancies that arise
when the desirability or undesirability
of events distorts the learning of their
probability of occurrence. The experi-
ments of Marks (60) and of Irwin (41),
cited earlier, are simplified examples of
the way in which desired outcomes in-
crease estimates of their likelihood of
occurrence. Certain more persistent
general personality tendencies also op-
erate in this sphere. It is indeed the
case that some people are readier to
expect and therefore quicker to perceive
the least desirable event among an array
of expected events, and others the most
desired. This is quite clearly a learned
adjustment to the events one is likely
to encounter, even if it may be sup-
ported by temperamental characteristics.
How such learning occurs, and why it is
so resistant to correction by exposure to
environmental events, are hardly clear.
But one matter that becomes increas-
ingly clear is that before we can know
much about how appropriate and inap-
propriate perceptual readiness is pro-
duced, we shall have to know much
more about how organisms learn the
probabilistic structure of their environ-
ments. This is a point that Brunswik
has made for some years ( 17), and it is
one that is now being taken seriously by
such students of probability learning as
Bush and Mosteller (18), Bruner, Good-
now, and Austin (9), Estes (23), Gal-
anter and Gerstenhaber (25), Hake and
Hyman (34), Edwards (22), and others.
There is another important feature of
learning that affects perceptual readi-
ness. It has to do with the range of
alternatives for which organisms learn
to be set perceptually. Put the matter
this way. It is a matter of common ob-
servation that some people are charac-
teristically tuned for a narrow range of
alternatives in the situations in which
they find themselves. If the environ-
ment is banal in the sense of containing
only high probability events and se-
quences or, more properly, events and
sequences that are strongly expected,
then the individual will do well and per-
ceive with a minimum of pause for close
looking. But should the environment
contain unexpected events, unusual se-
quences, then the result will be a marked
slowdown in identification and cate-
gorizing. Cue search must begin again.
We speak of such people as "rigid" or
"stuck." George Klein's work (46) on
shifting category judgments suggests
that, in general, people who are not able
to shift categorization under gradually
changing conditions of stimulation tend
also to show what he describes as "over-
control" on other cognitive and motiva-
tional tasks. At the other extreme is
specialization upon diversity, and how
such specialization is learned is equally
puzzling. I can perhaps best illustrate
the phenomenon by a commonly ob-
served pattern found in subjects in ta-
chistoscopic experiments. There are
subjects who show rather high thresholds
of identification generally, and who seem
to be "weighing" the stimulus in terms
of a wide array of interpretive cate-
gories. Jenkin ( 45) has recently de-
scribed such perception as "rational-
ON PERCEPTUAL READINESS 145
ized," the subject describing what he
sees as "like a so-and-so" rather than,
as in the "projective" response, report-
ing it "as a so-and-so." It is as if
the former type of response involved a
greater cue searching of stimulus inputs
for a fit to a wide range of things that
it "could be." It is also very likely
that premature sensory gating occurs in
individuals with a tendency to be set for
a minimum array of alternatives, lead-
ing them into error. The topic is one
that bears closer investigation. To any-
one who has had much experience in ob-
serving subjects in tachistoscopic work,
it seems intuitively evident that there
are large and individual differences pos-
sibly worth examining here.
We come finally to the vexing prob-
lem of "perceptual defense"-the man-
ner in which organisms utilize their per-
ceptual readiness to ward off events that
are threatening but about which there is
nothing they can do. There has been
foolish and some bitter ink spilled over
this topic, mostly because of a misunder-
standing. The notion of perceptual de-
fense does not require a little homuncular
ego, sitting behind a Judas-eye, capable
of ruling out any input that is potentially
disruptive-as even so able a critic as
F. H. Allport (1) seems to think. Any
preset filtering device can do all that is
required.
Let me begin with the general propo-
sition that failure to perceive is most
often not a lack of perceiving but a
matter of interference with perceiving.
Whence the interference? I would pro-
pose that the interference comes from
categorizations in highly accessible cate-
gories that serve to block alternative
categorizations in less accessible cate-
gories. As a highly speculative sugges-
tion, the mechanism that seems most
likely to mediate such interference is
probably the broadening of category ac-
ceptance limits when a high state of
readiness to perceive prevails; or, in the
language of the preceding section, the
range of inputs that will produce a
match signal for a category increases in
such a way that more accessible cate-
gories are likely to "capture" poor-fitting
sensory inputs. We have already con-
sidered some evidence for increase in ac-
ceptance limits under high readiness con-
ditions: the tendency to see a red four of
clubs as either a four of diamonds or a
four of clubs, with color-suit relation-
ship rectified ( 14), the difficulty of
spotting reversed letters imbedded in
the middle of a word ( 69), and so on.
Let us examine some experimental
evidence on the role of interference in
perceptual failure. Wyatt and Campbell
(86) have shown that if a subject de-
velops a wrong hypothesis about the
nature of what is being presented to
him for perception at suboptimal condi-
tions, the perception of the object in
terms of its conventional identity is
slowed down. This observation has
been repeated in other studies as well.
Postman and Bruner (68), for example,
have shown that if a subject is put under
pressure by the experimenter and given
to believe that he is operating below
standard, then he will develop prema-
ture hypotheses that interfere with cor-
rect perception of the word stimuli be-
ing presented to him. The authors refer
to "perceptual recklessness" as charac-
terizing the stressed subjects irt contrast
to those who operated under normal ex-
perimental conditions. It may well be,
just in passing, that stress has not only
the specific effect of leading to prema-
ture, interfering hypotheses but that it
disrupts the normal operation of match-
mismatch signaling systems in the nerv-
ous system. Unpublished studies from
our own laboratory carried out by
Bruner, Postman, and John (15) have
shown the manner in which subjects mis-
perceive low-probability contingencies
in terms of higher probability categories.
For example, a subject in the experi-
146 ]EROME s. BRUNER
mental group is shown tachistoscopically
a picture of a discus thrower, wound up
and ready to throw. In his balancing
arm and placed across the front of him
is a large bass viol. A control subject
is shown the same picture, the exact
space filled by the bass viol now being
occupied by the crouching figure of a
track official with his back to the camera.
The brightness, shading, and area of
the viol and the official are almost iden-
tical. Subjects begin by identifying the
first flash of the picture as an athlete
with a shadow across him. The subjects
faced with the incongruous picture then
go on with reasonable hypotheses-in-
cluding the hypothesis of a crouching
human figure, "probably an official," as
one subject put it-and in the process
of running through the gamut of likely
hypotheses, correct perception is inter-
fered with. It will not surprise you if I
report that the threshold for the incon-
gruous stimulus picture is markedly
higher than that for the more conven-
tional one.
Hypotheses and states of readiness
may interfere with correct perception in
yet another way: by creating a shifting
"noise" background that masks the cues
that might be used for identifying an
environmental event. At the common-
sense level this can best be illustrated
by reference to perceptual-motor learn-
ing where kinesthetic cues are of im-
portance. In teaching a person how to
cast a fly, it is necessary for him to
guide his forward delivery by feeling the
gentle pressure release that occurs when
the line reaches the end of its uncurving
on the backcast. If your flycasting
pupil is too eager to spot this cue, he
will be rather tense, and his own mus-
cular tension will mask the gentle pres-
sure release that he must use as a signal.
A good instance is provided by the
experiment of Goodnow and Pettigrew
(30) at Harvard. It is concerned with
the ability of subjects to perceive a regu-
larity in a sequence of events-a very
simple regularity, like the alternation
left-right-left-right .... The experiment
is done on a conventional two-armed
bandit, the subject having the task of
betting on whether a light will appear
on the left or on the right. The task is
simple. A subject is first given some
pretraining, in one of four pretraining
groups. One is given pretraining by
learning a simple alternation pattern of
payoff, another is trained to find the
payoff all on one side (not easy for all
subjects), a third is trained to find the
pattern LLRLLR ... , and a final group
is given no pretraining. Following the
pretraining and without pause, all sub-
jects are given a series of 60 choices in
which the payoff is randomly arranged,
the two sides totaling out to 50: 50. Im-
mediately following this random phase,
and again without pause, the payoffs
now go into a stage of simple alterna-
tion, LRLR. ... How long does it take
the subject to perceive the regularity of
the final temporal pattern? The speed
of discovery depends, it turns out, upon
the kinds of behavioral hypotheses a
subject develops during the phase of
random payoff. If he develops any
regularity of response-like win-stay-
lose-shift or win-shift-lose-stay-then he
will quickly spot the new pattern. Pre-
training on a constant one-side payoff
or on single alternation both produce
such regularity, and both forms of pre-
training produce equally good results-
the subject requiring but eight or nine
exposures to the pattern introduced after
the random phase to begin responding
without error. No pretraining, or pre-
training on the pattern LLRLLR . . . ,
does not produce the regularity of re-
sponse required. Instead, the subject
works on odd and constantly shifting
hypotheses during the random period.
When the single-alternation regularity is
introduced, the result is a marked re-
duction in ability to spot the new pat-
ON PERCEPTUAL READINESS 147
tern-some subjects failing to discover
the pattern in 200 trials. What we are
dealing with here is interference-hy-
potheses and responses serve to mask the
regularity of events in the environment.
In order for an environmental regu-
larity to be perceived, there has to be a
certain amount of steadiness in the hy-
potheses being employed and in the re-
sponse pattern that is controlled by it.
Short of this, masking and clumsy per-
ceptual performance results.
Now what has all this to do with "per-
ceptual defense"? The concept was in-
troduced some years ago by Postman
and myself as a description of the phe-
nomenon of failure to perceive andjor
report material known by independent
test to be regarded as inimical by the
subject. It was proposed ( 13) that
there was a hierarchy of thresholds, and
that an incoming stimulus could be re-
sponded to without its reaching the level
of reportable experience-as in the Mc-
Ginnies (58) and Lazarus and Mc-
Cleary (54) studies, where autonomic
response followed presentation of a po-
tentially traumatic stimulus without the
subject's being able to give a report of
the nature of the stimulus. The study
of Bricker and Chapanis (6) threw
further light on the concept of a hier-
archy of thresholds by demonstrating
that, though subjects could not report
spontaneously on the identity of the
shock syllables used by Lazarus and Mc-
Cleary, they could guess them well in
excess of chance if given a restricted
choice regarding what word had been
presented. I would like to propose two
additional factors that might lead to a
failure of perception of emotionally neg-
ative material.
It is conceivable that the estimates of
probability of occurrence of disvalued
events are, in some individuals, reduced
-essentially the obverse of what was
observed in the experiments of Marks
( 60) and Irwin ( 41), where probability
estimates were inflated by desirability.
If accessibility is decreased by such dis-
valuation, then a cognitive counterpart
of what is clinically called "repression"
can be posited. It is known, however,
that not everyone shows this tendency
to be unready for objects and events
that are anxiety-arousing. Others seem
to inflate their estimate of the likelihood
of occurrence of inimical events. Cer-
tainly one finds clinical evidence for
such a pattern among anxiety neurotics.
In an early paper, Postman and Bruner
(68) described two types of performance
with respect to known anxiety-produc-
ing stimuli, defense and vigilance, the
former a heightened threshold of iden-
tification for such stimuli, the latter a
lowered threshold. In a carefully de-
signed experiment contrasting the per-
formance of clinically diagnosed
11
intel-
lectualizers" and "repressors," Lazarus,
Eriksen, and Fonda (53) have shown
that the former group indeed are faster
in recognizing negatively charged mate-
rial than they are in recognizing neutral
material, while the latter show the re-
verse tendency. Again, I find it neces-
sary to revert to a point made earlier.
I do not think that we are going to get
much further ahead in understanding
hyper- and hyporeadiness for encounter-
ing anxiety-evoking stimuli short of do-
ing studies of the learning of environ-
mental probabilities for sequences con-
taining noxious and beneficial events.
One additional mechanism that may
be operative in lowering or generally in
altering readiness to perceive material
that in some way may be threatening. I
hesitate to speak of it in detail, since it
is of such a speculative order, and do so
only because some experiments suggest
themselves. It is this. Conceivably,
categories for classes of objects that are
pain-arousing are set up with narrow
acceptance limits for stimulus inputs re-
lated to them. That is to say, what we
speak of as "repression" may be the
148 jEROME S. BRUNER
establishment of very narrow category
limits that prevent the evocation of
match signals for inputs that do not fit
category specifications very precisely.
I am mindful that as far as autonomic
reactivity is concerned potentially trau-
matic stimuli work in quite the reverse
direction. If anything, a wide range of
objects, appropriate and inappropriate,
arouse autonomic reactions, without
leading to verbalizable report concern-
ing the categorial identity of the eliciting
objects. Yet it is conceivable that with
respect to one kind of threshold (auto-
nomic) the acceptance limits are broad
and with respect to another (
awareness) very narrow. I think it
would be worth while in any case to in-
vestigate the acceptance limits of inimi-
cal stimulus inputs by altering the char-
acteristics of objects so that, in essence,
one gets a generalization gradient for
recognition. My guess is that the gra-
dient will be much steeper for anxiety-
arousing stimuli than for neutral ones.
All that remains is to do the experiment.
Finally, it may also be the case that
category accessibility reflects the instru-
mental relevance of the environmental
events they represent. There is evidence
that the recognition threshold for noxious
objects about which one can do some-
thing is lower than normal, whereas for
ones about which nothing instrumental
can be done, the threshold is higher.
That is to say, words that signal a shock
that can be avoided show lowered thresh-
olds, words signaling unavoidable shock
show a threshold rise. One may well
speculate whether the instrumental rele-
vance of objects is not a controlling fac-
tor in guiding the kind of search be-
havior that affects category accessibility.
The problem needs much more thorough
investigation than it has received.
We have touched on various condi-
tions that might lead a person to be in-
appropriately set for the events he must
perceive easily and quickly in his en-
vironment. Many other studies could
be mentioned. But the intention has
not been to review the rather sprawling
literature in the field, but to propose
some possible mechanism affecting readi-
ness so that research might be given a
clearer theoretical direction.
CONCLUSIONS
We have been concerned in these
pages with a general view of perception
that depends upon the construction of a
set of organized categories in terms of
which stimulus inputs may be sorted,
given identity, and given more elabo-
rated, connotative meaning. Veridical
perception, it has been urged, depends
upon the construction of such category
systems, categories built upon the in-
ference of identity from cues or signs.
in fine, represents the range of
mferences about properties, uses, and
consequences that can be predicted from
the presence of certain criteria! cues.
Perceptual readiness refers to the rel-
ative accessibility of categories to af-
ferent stimulus inputs. The more ac-
cessible a category, the less the stimulus
input required for it to be sorted in
terms of the category, given a degree of
match between the characteristics of the
input and the specifications of the cate-
gory. In rough form, there appear to
be two general determinants of category
accessibility. One of them is the likeli-
hood of occurrence of events learned by
the person in the course of dealing with
the world of objects and events and the
redundant sequences in which these are
imbedded. If you will, the person builds
a model of the likelihood of events a
form of probability learning only ;ow
beginning to be understood. Again in
rough terms, one can think of this ac-
tivity as achieving a minimization of
surprise for the organism. A second de-
terminant of accessibility is the require-
ments of search dictated by need states
ON PERCEPTUAL READINESS 149
and the need to carry out habitual en-
terprises such as walking, reading, or
whatever it is that makes up the round
of daily, habitual life.
Failure to achieve a state of percep-
tual readiness that matches the proba-
bility of events in one's world can be
dealt with in one of two ways: either by
the relearning of categories and ex-
pectancies, or by constant close inspec-
tion of events and objects. Where the
latter alternative must be used, an or-
ganism is put in the position of losing
his lead time for adjusting quickly and
smoothly to events under varying condi-
tions of time pressure, risk, and limited
capacity. Readiness in the sense that
we are using it is not a luxury, but a
necessity for smooth adjustment.
The processes involved in "sorting"
sensory inputs to appropriate categories
involve cue utilization, varying from
sensorially "open" cue searching under
relative uncertainty, to selective search
for confirming cues under partial cer-
'tainty, to sensory "gating" and distor-
tion when an input has been categorized
beyond a certain level of certainty.
Four kinds of mechanisms are pro-
posed to deal with known phenomena
of perceptual categorizing and differ-
ential perceptual readiness: grouping
and integration, access ordering, match-
mismatch signal utilization, and gating.
The psychological evidence leading one
to infer such processes were examined
and possible neurological analogues con-
sidered. The processes are conceived
of as mediators of categorizing and its
forms of connectivity, the phenomena of
differential threshold levels for various
environmental events, the guidance of
cue search behavior, and lastly, the phe-
nomena of sensory inhibition and "filter-
ing."
Finally, we have considered some of
the ways in which failure of perceptual
readiness comes about-first, through a
failure to learn appropriate categories
for sorting the environment and for fol-
lowing its sequences, and second, through
a process of interference whereby more
accessible categories with wide accept-
ance limits serve to mask or prevent the
use of less accessible categories for the
coding of stimulus inputs. The concept
of "perceptual defense" may be re-ex-
amined in the light of these notions.
In conclusion, it seems appropriate to
say that the ten years of the so-called
New Look in perception research seem
to be coming to a close with much em-
pirical work accomplished-a great deal
of it demonstrational, to be sure, but
with a promise of a second ten years in
which hypotheses will be more rigor-
ously formulated and, conceivably, neu-
ral mechanisms postulated, if not dis-
covered. The prospects are anything
but discouraging.
REFERENCES
1. ALLPORT, F. H. Theories of perception
and the concept of structure. New
York: Wiley, 1955.
2. AscH, S. E. Social psychology. New
York: Prentice-Hall, 1952.
3. BARTLETT, F. C. Remembering. Cam-
bridge, England: Cambridge Univer.
Press, 1932.
4. BINDER, A. A statistical model for the
process of visual recognition. Psychol.
Rev., 1955, 62, 119-129.
5. BITTERMAN, M. E., & KNIFFIN, c. w.
Manifest anxiety and "perceptual de-
fense." J. abnorm. soc. Psychol., 1953,
48, 248-252.
6. BRICKER, P. D., & CHAPANIS
1
A. Do in-
correctly perceived tachistoscopic stimuli
convey some information? Psychol.
Rev., 1953, 60, 181-188.
7. BROWN, D. R. Stimulus similarity and
the anchoring of subjective scales.
Amer. J. Psychol., 1953, 66, 199-214.
8. BRUNER, J. S. Personality dynamics and
the process of perceiving. In R. R.
Blake & G. V. Ramsey (Eds.), Percep-
ception: an approach to personality.
New York: Ronald, 1951. Pp. 121-147.
9. BRUNER, J. S., GOODNOW, J. J., & AUSTIN,
G. A. A study of thinking. New
York: Wiley, 1956.
150 ]EROME S. BRUNER
10. BRUNER, J. s., MILLER, G. A., & ZIMMER-
MAN, C. Discriminative skill and dis-
criminative matching in perceptual rec-
ognition. J. exp. Psychol., 1955, 49,
187-192.
11. BRUNER, J. S., & MINTURN, A. L. Per-
ceptual identification and perceptual
organization. J. gen. Psychol., 1955,
53, 21-28.
12. BRUNER, J. S., & PosTMAN, L. Emotional
selectivity in perception and reaction.
J. Pers., 1947, 16, 69-77.
13. BRUNER, J. S., & POSTMAN, L, Percep-
tion, cognition, and behavior. J. Pers.,
1949, 18, 14-31.
14. BRUNER, J. S., & PosTMAN, L. On the
perception of incongruity: a paradigm.
J. Pers., 1949, 18, 206-223.
15. BRUNER, J. S., PosTMAN, L., & JoHN, W.
Normalization of incongruity. Research
memorandum. Cognition Project, Har-
vard Univer., 1949.
16. BRUNER, J. S., PosTMAN, L., & RoDRIGUEs,
J. Expectation and the perception of
color. Amer. J. Psychol., 1951, 64, 216-
227.
17. BRUNSWIK, E. Systematic and representa-
tive design of psychological experiments.
Berkeley: Univer. of California Press,
1949.
18. BusH, R. R., & MosTELLER, C. F. Sto-
chastic models for learning. New York:
Wiley, 1955.
19. CHAPMAN, D. W. Relative effects of de-
terminate and indeterminate Aufgaben.
Amer. J. Psychol., 1932, 44, 163-174.
20. CowEN, E. L., & BEIER, E. G. The in-
fluence of "threat expectancy" on per-
ception. J. Pers., 1951, 19, 85-94.
21. EccLES, J. C. The neurophysiological
basis of mind. Oxford: Oxford Univer.
Press, 1953.
22. EDWARDS, W. The theory of decision mak-
ing. Psychol. Bull., 1954, 51, 38G-417.
23. EsTES, W. K. Individual behavior in un-
certain situations: an interpretation in
terms of statistical association theory.
In R. M. Thrall, C. H. Coombs, &
R. L. Davis (Eds.), Decision processes.
New York: Wiley, 1954. Pp. 127-137.
24. FRY, D. P., & DENES, P. Mechanical
speech recognition. In W. Jackson
(Ed.), Communication theory. New
York: Academic Press, 1953.
25. GALANTER, E., & GERSTENHABER, M. On
thought: extrinsic theory of insight.
Amer. Psychologist, 1955, 10, 465.
26. GIBSON, J. J. The perception of the visual
world. Boston: Houghton Mifflin, 1950.
27. GmsoN, J. J., & GmsoN, E. J. Perceptual
learning: differentiation or enrichment?
Psychol. Rev., 1955, 62, 32-41.
28. GaLAMBos, R., SHEATz, G., & VERNIER, V.
G. Electrophysiological correlates of a
conditioned response in cats. Science,
1956, 123, 376-377.
29. GoooNow, J. J. Determinants of choice-
distribution in two-choice situations.
Amer. J. Psychol., 1955, 68, 106-116.
30. GooDNOW, J. J., & PETTIGREW, T. E. Some
difficulties in learning a simple pattern
of events. Paper presented at annual
meeting of the East. Psycho!. Ass., At-
lantic City, March, 1956.
31. GRAHAM, C. H. Perception and behavior.
Presidential address to the East. Psy-
cho!. Ass., Atlantic City, March, 1956.
32. GRANIT, R. Receptors and sensory per-
ception. New Haven: Yale Univer.
Press, 1955.
33. HAIRE, M., & GRUNES, W. F. Perceptual
defenses: processes protecting an organ-
ized perception of another personality.
Hum. Relat., 1950, 3, 403-412.
34. HAKE, H. W., & HYMAN, R. Perception of
the statistical structure of a random
series of binary symbols. J. exp. Psy-
chol., 1953, 45, 64-74.
35. HARPER, R. S., & BoRING, E. G. Cue&.
Amer. J. Psychol., 1948, 61, 119-123.
36. HEBB, D. 0. The organization of behavior.
New York: Wiley, 1949.
37. HELSON, H. Adaptation-level as a basis
for a quantitative theory of frames of
reference. Psychol. Rev., 1948, 55, 297-
313.
38. HERNANDEZ-PEoN, R., ScHERRER, R. H., &
JoUVET, M. Modification of electric ac-
tivity in the cochlear nucleus during
"attention" in unanesthetized cats.
Science, 1956, 123, 331-332.
39. HoRNBOSTEL, E. M. VON. Unity of the
senses. Psyche., 1926, 7, 83-89.
40. HowEs, D. On the interpretation of word
frequency as a variable affecting speed
of recognition. J. exp. Psycho!., 1954,
48, 106-112.
41. IRWIN, F. W. Stated expectations as func-
tions of probability and desirability of
outcomes. J. Pers., 1953, 21, 329-335.
42. ITTLESON, W. H. The Ames demonstra-
tions in perception. Princeton, N. J.:
Princeton Univer. Press, 1952.
43. JARRETT, J. Strategies in risk-taking situa-
tions. Unpublished doctor's dissertation,
Harvard Univer. Library, 1951.
44. ]ARVIK, M. E. Probability learning and a
negative recency effect in the serial an-
ON PERCEPTUAL READINESS 151
ticipation of alternative symbols. J.
exp. Psychol., 1951, 41, 291-297.
45. JENKIN, N. Two types of perceptual ex-
perience. J. clin. Psychol., 1956, 12,
44-49.
46. KLEIN, G. S. The personal world through
perception. In R. R. Blake & G. V.
Ramsey (Eds.), Perception: an ap-
proach to personality. New York:
Ronald, 1951. Pp. 328-355.
47. KOHLER, I. Rehabituation in perception.
Published separately in three parts, in
German, in Die Pyramide, 1953, Heft
5, 6, and 7 (Austria). Translated by
Henry Gleitman and edited by J. J.
Gibson. Privately circulated by the
editor.
48. KoHLER, W. Dynamics in psychology.
New York: Liveright, 1940.
49. KuFFLER, S. W., HuNT, C. C., & QuiLLIAN,
]. P. Function of medullated small-
nerve fibers in mammalian ventral roots:
efferent muscle spindle innervation. J.
Neurophysiol., 1951, 14, 29-54.
50. KUFFLER, S. W., & HUNT, C. C. The
mammalian small nerve fibers: a sys-
tem for efferent nervous regulation of
muscle spindle discharge. Proc. Assoc.
Res. Nerv. Ment. Dis., 1952, Vol. 30.
51. LASHLEY, K. S. Experimental analysis of
instinctive behavior. Psychol. Rev.,
193 8, 45, 445-4 71.
52. LASHLEY, K. S. In search of the engram.
Symp. Soc. Exp. Biol., 1950, 4, 454-
482.
53. LAZARUS, R. S., ERIKSEN, c. w., & FONDA,
C. P. Personality dynamics and audi-
tory perceptual recognition. J. Pers.,
1951, 19, 471-482.
54. LAZARus, R. S., & McCLEARY, R. A. Au-
tonomic discrimination without aware-
ness: a study of subception. Psychol.
Rev., 1951, 58, 113-222.
55. LEKSELL, L. The action potential and
excitatory effects of the small ventral
root fibers to skeletal muscles. Acta
Physiol. Scand., 1945, 10, Suppl. 31.
56. LENNEBERG, E. H. An empirical investiga-
tion into the relationship between lan-
guage and cognition. Unpublished doc-
toral dissertation, Harvard Univer. Li-
brary, 1956.
57. LORENTE DE No, R. Transmission of im-
pulses through cranial motor nuclei. J.
Neurophysiol., 1939, 2, 402-464.
58. McGINNIES, E. Emotionality and per-
ceptual defense. Psychol. Rev., 1949,
56, 244-251.
59. MAcKAY, D. M. Toward an information-
flow model of human behavior. Brit.
J. Psychol., 1956, 47, 30-43.
60. MARKS, R. W. The effect of probability,
desirability, and "privilege" on the state
of expectations of children. J. Pers.,
1951, 19, 332-351.
61. MILLER, G. A., BRUNER, J. s., & POSTMAN,
L. Familiarity of letter sequences and
tachistoscopic identification. J. gen.
Psychol., 1954, 50, 129-139.
62. MILLER, G. A., HEISE, G. A., & LICHTEN,
W. The intelligibility of speech as a
function of the context of the test ma-
terials. J. exp. Psychol., 1951, 41, 329-
335.
63. PEIRCE, C. S. How to make our ideas
clear. Popular Sci. Mon., 1878, 12, 286-
302.
64. PENFIELD, W. Memory mechanisms. Arch.
Neurol. & Psychiat., 1952, 67, 178-191.
65. PIAGET, J, Play, dreams, and imitation in
childhood. New York: Norton, 1951.
66. PRATT, C. C. The role of past experience
in visual perception. J. Psycho/., 1950,
30, 85-107.
67. PRENTICE, W. C. H. Paper read at the
Symposium on Conceptual Trends in
Psychology, at Amer. Psycho!. Ass., New
York, September, 1954.
68. POSTMAN, L., & BRUNER, J, S. Perception
under stress. Psychol. Rev., 1948, 55,
314-323.
69. POSTMAN, L., BRUNER, J. S., & wALK, R.
D. The perception of error. Brit. J.
Psychol., 1951, 42, 1-10.
70. POSTMAN, L., & CRUTCHFIELD, R. S. The
interaction of need, set, and stimulus
structure in a cognitive task. Amer. J.
Psychol., 1952, 65, 196-217.
71. SELFRIDGE, 0. Pattern recognition and
learning. Memorandum of Lincoln
Laboratory, Massachusetts Institute of
Technology, 1955.
72. SoLOMON, R. L., & PosTMAN, L. Fre-
quency of usage as a determinant of
recognition thresholds for words. J.
exp. Psychol., 1952, 43, 195-201.
73. SMITH, J. W., & KLEIN, G. S. Cognitive
control in serial behavior patterns. Dit-
toed manuscript, available from author,
1951.
74. STEVENS, S. S. Chapter I in S. S. Stevens
(Ed.), Handbook of experimental psy-
chology. New York: Wiley, 1951.
75. STEVENS, S. S. The direct estimation of
sensory magnitudes-loudness. Amer.
J. Psychol., 1956, 69, 1-25.
152 JEROME s. BRUNER
76. TANNER, W. P., JR., & SwETS, J, A. A
decision-making theory of human de-
tection. Psychol. Rev., 1954, 61, 401-
409.
77. TINBEROEN, N. The study of instinct.
Oxford: Oxford Univer. Press, 1951.
78. TITCHENER, E. B. A beginner's psy-
chology. New York: Macmillan, 1916.
79. ToLMAN, E. C. Discussion. l. Pers.,
1949, 18, 48-50.
80. UTTLEY, A. M. The conditional proba-
bility of signals in the nervous system.
Radar Research Establ., British Minis-
try of Supply, Feb., 1955.
81. VERNON, M. D. A further study of visual
perception. Cambridge, England: Cam-
bridge Univer. Press, 1952.
82. VoLKMANN, J. In M. Sherif & J. H.
Rohrer (Eds.), Social psychology at the
crossroads. New York: Harpers, 1951.
83. WALLACH, H. Some considerations con-
cerning the relation between perception
and cognition. J. Pers., 1949, 18, 6-13.
84. WHITE, M. The age of analysis. New
York: New American Library, 1955.
85. WooDWORTH, R. S. Reenforcement of per-
ception. Amer. l. Psychol., 1947, 60,
119-124.
86. WYATT, D. F., & CAMPBELL, D .. T. On the
liability of stereotype or hypothesis.
J. abnorm. soc. Psychol., 1951, 46, 496-
500.
87. YOKOYAMA, J. Reported in E. G. Boring,
A history of experimental psychology.
(2nd Ed.). New York: Appleton-Cen-
tury, 1954.
88. YouNo, J. Z. Doubt and certainty in
science. Oxford: Oxford Univer. Press,
1951.
(Received June 4, 1956)